Media assets, for example, movies or television (TV) programming, are typically rich in features and may require a variety of different categories of metadata to adequately describe their content. However, the conventional generation of metadata descriptive of a particular media asset is inadequate for enabling an effective comparison of content features across a library of such media assets. As a result, the conventional approach to generating and storing metadata descriptive of media assets fail to address large scale searching and cross-referencing of those media assets.